Director, Molecular AI & Federated Learning

Scorpion Therapeutics; Boston (MA); On-site; USD 180,000 - 240,000; Full time

Job summary

Scorpion Therapeutics is seeking a Director of Molecular AI & Federated Learning to define the technical vision for privacy-preserving federated learning and generative small-molecule design. The role leads through vision, rigor, and mentorship while not managing people directly.

You will architect federated foundation models, advance self-supervised learning, optimize scalability, and drive multi-task molecular property predictions, with emphasis on ADMET, solubility, and synthetic feasibility.

Qualifications

• PhD in a relevant field (CS/ML/chemistry or related).
• 5+ years of ML experience in drug discovery or equivalent leadership impact.

Responsibilities

• Define and set the technical direction for federated molecular AI and foundation models.
• Mentor and act as principal technical authority; review methods and code.
• Architect federated foundation models for distributed partner data.
• Advance semi-/self-supervised learning for federated constraints.
• Develop communication-efficient federated optimization techniques for non-IID data.
• Profile and optimize scalability; build simulation environments.
• Design multi-task learning with personalized models and meta-learning.
• Create protocols for downstream adaptation and task-specific validation.
• Develop multi-task small-molecule property prediction models (ADMET, solubility, permeability).
• Design/deploy generative chemistry models for de novo design and lead optimization.
• Implement multi-objective design with Pareto-front exploration.

Skills

Federated learning

Distributed ML

Molecular AI

Graph neural networks

Generative chemistry

Leadership (non-manager)

Interpretability/XAI

Publication record

Education

PhD in CS/ML/Computational Chemistry

Tools

PyTorch

RDKit/DeepChem

Scikit-learn

Git

Job description

Job Summary

Director, Molecular AI & Federated Learning (TuneLab) — Senior technical leadership role setting the technical vision for privacy-preserving federated learning and generative small-molecule design; leads through vision, rigor, and mentorship (not formal people management).

Key Responsibilities

• Set technical direction and research agenda for federated molecular AI (foundation models, multi-task learning, generative small-molecule design).
• Mentor and serve as principal technical authority; guide experimental design and review methods/code.
• Architect federated foundation models (e.g., Transformers, graph neural networks) for pre-training across distributed partner data.
• Advance semi-/self-supervised learning suited to federated constraints (communication bottlenecks, data heterogeneity).
• Develop communication-efficient federated optimization/aggregation (FedAvg, FedProx, SCAFFOLD) for non-IID data.
• Profile/optimize scalability (memory, latency, communication cost) and build simulation environments.
• Architect federated multi-task learning and handle task/feature heterogeneity (personalized models, meta-learning, gradient aggregation; prevent negative transfer).
• Create protocols for downstream adaptation/validation with per-task metrics and fairness assessment.
• Build multi-task small-molecule property prediction (ADMET, solubility, permeability, metabolic stability, off-target liabilities).
• Design/deploy generative chemistry models (VAEs, diffusion, flow matching, autoregressive) for de novo design, lead optimization, scaffold hopping.
• Implement integrated ADMET-driven, multi-objective design (Pareto-front exploration) and synthetic feasibility exploration.
• Apply interpretability/XAI and establish rigorous benchmarking and reproducible governance (public + proprietary data; publications/presentations).

Basic Qualifications

• PhD in CS, Computational Chemistry, Cheminformatics, ML, Computational Biology, or related field.
• 5+ years post-PhD ML experience in drug discovery (preference for 8+ years) or equivalent technical leadership/impact.

Additional Preferences

• Technical leadership and mentoring without formal people-management.
• Generative molecular design; multi-task/representation-learning track record; medicinal chemistry + ADMET optimization.
• Hands-on federated learning/distributed optimization/privacy-preserving ML; top-tier publications.
• Graph/geometric deep learning; organic chemistry + synthetic feasibility; fragment/structure-based design.
• PK/PD modeling and clinical translation; RDKit/DeepChem and PyTorch; active learning and D-M-T-A cycles.
• Uncertainty quantification and XAI in federated/multi-task settings; exceptional cross-disciplinary communication; independent, self-directed.

Other Information

• Location: Indianapolis, San Francisco, or Boston; up to 10% travel; attendance expected at key conferences.